RLIPS
RLIPS (R Linear Inverse Problem Solver) is a R package for solving large overdetermined (stochastic) linear inverse problems. It is a modification of FLIPS (Fortran Linear Inverse Problem Solver) written in C and OpenCL.
As in FLIPS, RLIPS transforms the original linear system into a simple upper triangular one by using Givens rotations. However, in RLIPS these rotations are made in parallel using OpenCL and GPU's (Graphics Processing Units).
RLIPS is developed as part of EISCAT_3D project.
RLIPS is licensed under freeBSD License.
Copyright (c) University of Oulu, Finland
Written by Mikko Orispää
Download
rlips_0.5-3.tar.gz, Source package, 2013-03-21.
Version 0.5-3: Bug fixes.
Installation
Prerequisites
- R, version 2.10 or newer
- OpenCL capable graphics card or other GPU
- OpenCL implementation including drivers and OpenCL libraries
All Apple computers with OS X.6 (Snow Leopard) or newer have OpenCL capable graphics cards and preinstalled OpenCL framework.
For Linux, the current version of RLIPS supports only NVIDIA's OpenCL implementation which is bundled with NVIDIA's CUDA. Installation using other OpenCL implementations is possible but it requires manual modifications to the installation files (See below).
R package installation
For Apple computers and Linux systems with NVIDIA OpenCL installed, RLIPS is installed in the standard way. For example, using Terminal:
sudo R CMD INSTALL rlips_0.5-2.tar.gz
On Linux systems with another OpenCL implementation installed it might be necessary to untar the RLIPS package
tar xzf rlips_0.5-2.tar.gzand modify the file
rlips/src/Makevarsso that the compiler finds the OpenCL header and library files. We are working on a better installation script, so this inconvenience will disappear in the future.
Usage
In R, RLIPS is loaded in the standard way
require(rlips)
Help can be obtained by
help(rlips)Help files will also have examples of use (in the near future).
Questions, comments & bug findings
If there are any questions or comments, or you find bugs or errors, please do not hesitate to contact me.